Skip to content
Search

THE RADAR

What shipped this week

What the companies and projects we follow have released, in plain text and with a link to the original announcement. This is where the classes come from: when a change is worth building on, the class shows up here too.

24 sources watched

  • Google Research
  • Qiskit
  • Quanta Magazine
  • The Quantum Insider
  • Activepieces
  • n8n
  • n8n
  • Zapier
  • Claude Platform
  • Browser Use
  • Claude Code
  • CrewAI
  • Google DeepMind
  • Google ADK
  • LangGraph
  • MCP TypeScript SDK
  • OpenAI Agents SDK
  • OpenAI
  • SDK de OpenAI
  • Vercel AI SDK
  • Google
  • Hugging Face
  • MIT Technology Review
  • NVIDIA

  • Hugging Face TransformersSageminor releaseQueued for a class

    Hugging Face Transformers 5.20

    Special surprise including the new model Beam from Reflection AI :partyingface:. <img width="2912" height="1440" alt="GeminiGeneratedImageox4hjlox4hjlox4h" src=" />. Beam is a token-efficient, frontier-scale model focused on coding and agentic use cases. It is built for teams that want to run a powerful model on their own hardware, infrastructure, or cloud, and customize it to their specific needs. Links: Documentation. Fix small bug in auto mappings.

    Read the article →Read the original announcement

  • Browser UseNovaNo class

    Browser Use 0.13

    This release includes the OpenAI browser-tool driver and fixes two integration issues. Handle temporary empty Cloud tab URLs during navigation. Keep screenshots, crops, and pointer coordinates aligned on scaled displays. Provide 18 default browser tools and six optional actions through Browser Use Actor. Support local browsers, Browser Use Cloud, and borrowed browser sessions. Install the openai-tools extra with OpenAI Python 3.31+. OpenAI 3.x uses httpx2;

    Read the article →Read the original announcement
  • VercelCipherfirst sightingQueued for a class

    Vercel Ship 2026 recap

    For a decade, Vercel has shaped how the web is built. Now, we’re doing the same for agents. The companies that win the next decade will build on infrastructure designed for agents from the start, and thousands of people gathered across London, Berlin, New York, Sydney, and San Francisco to do just that at Vercel Ship 2026.

    Read the article →Read the original announcement

  • Claude PlatformNovafirst sightingQueued for a class

    Claude Platform

    We've launched OpenTelemetry export in beta, so you can trace Messages API requests and Claude Managed Agents sessions in your observability backend. The Claude SDK emits spans in Python, TypeScript, C#, Go, Java, PHP, and Ruby. To get spans from the API, add an OpenTelemetry destination in the Claude Console. Claude Managed Agents scheduled deployments are now available on Claude Platform on AWS .

    Read the article →Read the original announcement

  • Google ADKNovafirst sightingQueued for a class

    Google ADK 2.12

    This release enhances execution performance in long sessions while introducing finer control over tool execution, workflow approvals, and deployment security. Performance: Cut time to first model request by 40% in long sessions. Tools: Intercept and filter tool outputs before returning them to the model using aftertoolcallback. Workflows: Require human confirmation before executing workflow nodes via requireconfirmation.

    Read the article →Read the original announcement

  • Playwrightforgeminor releaseQueued for a class

    Playwright 1.64

    New page.webmcp and frame.webmcp give access to the tools that a page registers through the experimental WebMCP browser API, so you can test them like any other part of your app:. const browser = await chromium.launch({ args: ['--enable-features=WebMCP'] });. const page = await browser.newPage();. await page.goto('. for (const tool of await page.webmcp.tools()). console.log(tool.name, tool.description);. const result = await page.webmcp.callTool('add', { a: 2, b: 40 });.

    Read the article →Read the original announcement

  • Hugging Face TransformersSageminor releaseQueued for a class

    Hugging Face Transformers 5.19

    <img width="2716" height="2308" alt="image" src=" />. EmbeddingGemma 2 is a multimodal embedding model from Google built on the Gemma 4 architecture. It encodes text, images, audio, and video, individually or combined in one input, into a shared 768-dimensional vector space for cross-modal retrieval, semantic similarity, clustering, and classification. It uses Matryoshka Representation Learning, so embeddings can be truncated to 512, 256, or 128 dimensions.

    Read the article →Read the original announcement

  • Hugging Face TransformersSageminor releaseQueued for a class

    Hugging Face Transformers 5.18

    <img width="1680" height="900" alt="image" src=" />. Nemotron 3 Diarization is an open-weight streaming speaker diarization model designed to determine "who spoke when" in real-world audio. It supports both streaming and offline inference, handles up to eight speakers, and orders speaker outputs by each speaker's first arrival in the input audio. The model uses the Arrival-Order Speaker Cache (AOSC) 1 and FIFO queue introduced for Streaming Sortformer 1, 2.

    Read the article →Read the original announcement

  • OllamaNovaminor releaseQueued for a class

    Ollama 0.35

    Ollama now supports decision models through /v1/systemone, based on TypeSafe’s Jev API. Decision models return choices, probabilities, and scores instead of text. Use them for tasks such as ticket triage, model routing, and content classification. Available models:. Nimble from Bespoke Labs. Tev1 from Together AI. ollama pull nimble. Send context and one or more questions:. -H 'Content-Type: application/json' \. "model": "nimble",. "state": "Our checkout has returned 500 errors since 9am.",.

    Read the article →Read the original announcement

  • ActivepiecesFlowNo class

    Activepieces 0.92

    refactor(web): variable fonts, and a real --font-mono for the first time @moaaz-ae. feat(delay): improve the step forms @Talaljaber. feat(analytics): stamp marketing attribution on the PostHog person at sign-up @moaaz-ae. change the image download screenshot in builder canvas control [@mintlify[bot]]. chore(deps): bump dependencies @KhaledR57. feat: add deno code sandbox @MrChaker. feat(shopify): add Dev Dashboard client-credentials auth alongside Admin Token @OdaiAhmed99.

    Read the article →Read the original announcement
  • AstropixelNo class

    Astro 7.3

    #17736 2b8b2e8 Thanks @ematipico! - Adds a new container function called renderComponent(), which renders Astro components with inlined styles and scripts. Users must import the component with the new ?container query string:. import { experimentalAstroContainer } from "astro/container";. import TodoList from "./components/TodoList.astro?container";. const container = await experimentalAstroContainer.create();. const string = container.renderComponent(TodoList);

    Read the article →Read the original announcement
  • VercelCipherfirst sightingNo class

    Vercel Connect now supports TanStack AI

    Agents built with TanStack AI can now call OAuth-protected MCP servers through Vercel Connect, with no credentials for you to store or rotate. The new @vercel/connect/tanstack-ai subpath exports connectMCPTransport , which takes a TanStack transport config and attaches a Connect-backed auth provider. The provider is called before every MCP request, so the token is always fresh. If the user has not granted access, createMCPClient fails with a consent challenge before the model runs.

    Read the article →Read the original announcement

  • Claude CodeNovaNo class

    Claude Code 2.1

    Added Claude Opus 5.5 (claude-opus-5-5), now the default Opus model — 1M context, $4/$20 per Mtok with $0.20/Mtok cache reads. Added mouse support to more lists in fullscreen mode: the wheel scrolls the /skills list, and a skill's state options in /plugin can be clicked. Added CLAUDECODEMAXMCPDESCRIPTIONLENGTH to change the 2,048-character cap on MCP tool descriptions and server instructions for every MCP server in the session.

    Read the article →Read the original announcement
  • Node.jskernelNo class

    Node.js 26.10

    \[c0a42d23e5] - (SEMVER-MINOR) crypto: add crypto.parsePKCS12() (Brian Muenzenmeyer) #65627. \[eb4fabe81e] - doc: add araujogui to collaborators (Guilherme Araújo) #66090. \[64fb33d791] - (SEMVER-MINOR) ffi: load libraries from a mounted VFS (Matteo Collina) #65909. \[2b1701f810] - (SEMVER-MINOR) fs: add openAsBlobSync (greenhead) #65644. \[080e76b3d7] - (SEMVER-MINOR) net: support sending net.BoundSocket to threads and child processes (Guy Bedford) #64725.

    Read the article →Read the original announcement
  • vLLMSageNo class

    vLLM 0.30

    This release features 762 commits from 315 contributors (104 new)!. New models: DeepSeek-V4.1-Flash (#56214, #56228, #56208) with the whole KV stored in MXFP8 through the FlashMLA V4.1 record on SM100 , DeepGEMM Mega-mHC , and async Engram prefetch with Engram DP sharding ; DeepSeek-V4-Flash-Vision-Exp , also on ROCm and with LoRA ; GLM-5.3-Flash with EPLB ; K2-Horizon ; Cohere Compass ; Bailing V3 VL ; Nanbeige4.2 via the Transformers backend ;

    Read the article →Read the original announcement
  • Claude PlatformNovafirst sightingNo class

    Claude Platform

    We've launched Claude Opus 5.5 ( claude-opus-5-5 ), a model for long-running agentic coding and knowledge work. It has a 1M token context window by default, 128k max output tokens, and always-on adaptive thinking , at $4 / $20 USD per MTok (Claude Opus 5 is $5 / $25). Claude Opus 5.5 is available on the Claude API, Claude in Amazon Bedrock , Claude Platform on AWS , Claude on Google Cloud , and Claude in Microsoft Foundry .

    Read the article →Read the original announcement
  • VercelCipherfirst sightingNo class

    Vercel: GPT-6 Sol and Luna now available on AI Gateway

    GPT-6 Sol and GPT-6 Luna from OpenAI are now available on AI Gateway . Both models bring GPT-6 improvements in professional work, coding, computer use, factuality, and communication at a lower price than GPT-6 Astra . GPT-6 Sol ( openai/gpt-6-sol ) is suited to complex professional workflows and sustained coding tasks where quality and room to iterate both matter. GPT-6 Luna ( openai/gpt-6-luna ) is the lower-cost option for high-volume agentic workflows, coding, and everyday tasks.

    Read the article →Read the original announcement

  • VercelCipherfirst sightingNo class

    Vercel Connect now supports Microsoft Teams

    Vercel Connect now includes a managed connector for Microsoft Teams . Creating one gives your organization a Teams bot that your apps and agents run. People can @mention it in channels or message it directly, and your code receives the message and replies as the bot. As a Vercel Managed Connector , Vercel registers the Entra app and Azure Bot resource in your tenant, so there's no client secret to store. A tenant administrator with an Azure subscription completes setup once.

    Read the article →Read the original announcement
  • Drizzle ORMatlasNo class

    Drizzle ORM 0.45

    Installation:. npm i @netlify/db. Usage example:. import { drizzle } from 'drizzle-orm/netlify-db';. // reads NETLIFYDBURL and NETLIFYDBDRIVER env vars. const db = drizzle();. const result = await db.execute('select 1');. import { drizzle } from 'drizzle-orm/netlify-db';. const db = drizzle(process.env.DATABASEURL);. const result = await db.execute('select 1');. import { drizzle } from 'drizzle-orm/netlify-db';. // Explicit client — consumer controls the driver.

    Read the article →Read the original announcement

  • SDK de OpenAINovaNo class

    SDK de OpenAI 7.20

    api: add environment-variable vault credentials. api: add external storage configuration management. api: add safety case retrieval. api: add safety warning and deactivation webhook events. api: add SIP media security to incoming call events. api: Preserve request options in legacy GET calls

    Read the article →Read the original announcement
  • OllamaNovaNo class

    Ollama 0.34

    GET /api/show now advertises each model's thinking controls and default:. Available in the CLI with:. ollama show gemma4. levels false, true. Available in the API with:. curl -d '{"model": "glm-5.3-flash:cloud"}'. "thinking": {. "values": ["low", "high", "max"],. "default": "max". Also available on ollama.com directly for cloud models. Nemotron H vision models are now supported on Apple Silicon with MLX. Ollama's macOS app will now no longer reopen windows you've closed when activating it.

    Read the article →Read the original announcement

  • VercelCipherfirst sightingQueued for a class

    Vercel: Sub-second artifact deployments are now supported in Vercel CLI

    You and your agents can now deploy static artifacts to Vercel in under one second through Vercel CLI. Run vercel deploy to share a prototype, publish an HTML report, or preview a page created by your coding agent. Vercel automatically detects eligible deployments, and valid artifacts will skip the build step, immediately returning a live URL.

    Read the article →Read the original announcement
  • VercelCipherfirst sightingQueued for a class

    Vercel: Run Terminal-Bench and other Harbor evals on Vercel Sandbox

    You can now run Harbor evals on Vercel Sandbox. Harbor is the open-source harness behind Terminal-Bench , whose registry includes many other benchmarks such as SWE-bench, tau3-bench and OSWorld. Pass --env vercel to harbor run and each trial executes in its own isolated Firecracker microVM, so you can parallelize far beyond what your local machine is capable of. A task's network policy is enforced at the sandbox firewall, outside the VM.

    Read the article →Read the original announcement
  • CloudflareCipherfirst sightingNo class

    Cloudflare: Workflows, Workers - Delete Workflow instances individually or in batches

    You can now delete one or up to 100 Workflow instances and their stored state via the Workflows API or Wrangler 4.125.0 and later. Deleting an instance frees its stored state and stops its current execution. Storage billing is based on the average daily peak. Delete one instance by calling delete() on its handle: const instance = await env. MYWORKFLOW . get ( "instance-abc" ); await instance. delete (); If a Workflow deletes its own instance, execution stops during await instance.delete() .

    Read the article →Read the original announcement
  • NetlifyCipherfirst sightingNo class

    Netlify: TypeSafe Jev now available in AI Gateway

    TypeSafe’s Jev model is now available through Netlify’s AI Gateway with zero configuration required. Install @typesafe-ai/sdk and use it directly in your Netlify Functions — no API keys to create, no provider config, no base URLs to wire up. AI Gateway handles credentials automatically, and usage is billed to your Netlify credits like every other model in the gateway. Jev is TypeSafe’s first “System One” model, and it works differently from the chat models you’re used to.

    Read the article →Read the original announcement

  • SDK de OpenAINovaNo class

    SDK de OpenAI 7.17

    api: add compaction progress events. ci: retry transient Deno download failures. preserve chat runner abort reasons. realtime: preserve native WebSocket error causes. responses: reject function-based API keys in WebSocket constructors. runners: reject unfinished turns in streaming runTools. streaming: emit terminal SSE events missing a trailing blank line. zod: omit impossible optional branch in strict schemas. update lint tooling and tailor anti-slop for the SDK.

    Read the article →Read the original announcement
  • Vercel AI SDKNovapatchNo class

    Vercel AI SDK 7

    91c2128: feat(ai): add mid conversation tool discovery/updates for code-mode. 25a0447: feat(ai): deprecate rawInput in output-error UI message parts. 2cd80b3: Keep default Node.js downloads protected by DNS validation and connection pinning when frameworks or instrumentation wrap global fetch before or after the SDK loads. 123d71f: Add experimentalevaluate and the isolated experimental v4 evaluation model specification for Choice, Score, and Boolean questions against shared state.

    Read the article →Read the original announcement
  • Node.jskernelNo class

    Node.js 26.9

    \[d414624dce] - (SEMVER-MINOR) crypto: add a generic MAC API (Filip Skokan) #65553. \[7ac458f802] - (SEMVER-MINOR) crypto: discover ciphers from OpenSSL providers (Filip Skokan) #65484. \[7c7e95a3bd] - (SEMVER-MINOR) crypto: discover hashes from OpenSSL providers (Filip Skokan) #65484. \[2d2f4f6d1a] - (SEMVER-MINOR) ffi: enable module by default (Matteo Collina) #65475. \[657415c6df] - (SEMVER-MINOR) lib: implement node:bench (James M Snell) #65606.

    Read the article →Read the original announcement
  • BiomeforgeNo class

    Biome 2.5

    #9022 0d49e24 Thanks @dyc3! - Added the nursery rule noReturnInFinally. This rule disallows return statements in Promise.prototype.finally() callbacks, including inside nested blocks and conditional branches. Returns in nested functions are ignored by the rule. // Invalid: return in finally callback. Promise.resolve(1).finally(() => { return 2 }). // Valid: no return in finally callback. Promise.resolve(1).finally(() => { console.log(2) }).

    Read the article →Read the original announcement
  • VercelCipherfirst sightingNo class

    Vercel: Is Agentic now tailors its audit by site type

    Is Agentic reports now let you view your checks through one of four site types: Docs & content, Business, App, or Commerce. For example, the Commerce view highlights payment and checkout standards like x402, UCP, and ACP, while the App view highlights API discovery, authentication, error handling, and SDK support. Your score stays the same in every view, and you can change views on an existing report without rescanning.

    Read the article →Read the original announcement

  • Vercel AI SDKNovaNo class

    Vercel AI SDK 7

    5c0054d: Add optional browser-direct WebRTC for experimental client-delegated Live conversations alongside the existing WebSocket path. Exchange SDP through an application endpoint with api.session, configure server-owned data-channel permissions, and preserve committed React session ownership. Capture follows the selected sender track, borrowed tracks remain caller-owned, and disconnect recovery and finalization stay bounded.

    Read the article →Read the original announcement
  • CloudflareCipherfirst sightingNo class

    Cloudflare: Workflows, Workers - Stream Workflow instance events in your Worker or via the API with .subscribe()

    You can now stream Workflow instance events via WorkflowInstance.subscribe() and the GET /subscribe API endpoint. Workers and HTTP clients can react to workflow and step events, including attempts, sleeps, waits, and rollbacks, without polling for instance status. A subscription first streams the entire event history of the Workflow instance. After streaming past events, the subscription waits for new events as the instance runs.

    Read the article →Read the original announcement
  • VercelCipherfirst sightingQueued for a class

    Vercel 3.8

    Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking from Google are now available on AI Gateway. Both models support real-time spoken interactions for voice assistants, conversational experiences, and applications that respond through audio. google/gemini-3.8-live supports real-time audio, visual grounding, automatic switching across 97 languages, and background tool calls while the conversation continues.

    Read the article →Read the original announcement
  • NetlifyCipherfirst sightingNo class

    Netlify 4.1

    DeepSeek V4.1 Flash is now available through OpenRouter on Netlify’s AI Gateway with zero configuration required. Use the OpenRouter SDK directly in your Netlify Functions without managing API keys or authentication. AI Gateway handles everything automatically. Here’s an example using DeepSeek V4.1 Flash: import { OpenRouter } from '@openrouter/sdk'; export default async () => { const client = new OpenRouter();

    Read the article →Read the original announcement

  • OllamaNovaNo class

    Ollama 0.34

    MLX safetensors ollama create no longer experimental. GGUF model creation now requires using llama.cpp tooling for safetensor conversion and quantization. Improved MLX memory handling on Apple Silicon. Runaway repeat token detection now requires 100 repeat tokens for reduced false positives (e.g. OCR). /api/tags is much faster on large model libraries (3.1 s → 294 ms cold in testing), and model capabilities are now reported consistently.

    Read the article →Read the original announcement
  • Claude CodeNovapatchNo class

    Claude Code 2.1

    Added fast mode in Claude Code Remote sessions (cloud and self-hosted runners): the host's fast-mode setting or /fast typed in the session applies where your organization allows it. Added mouse support to the /config panel in fullscreen mode: the wheel scrolls the settings list, a click on a setting's value changes it, and the row under the pointer is highlighted.

    Read the article →Read the original announcement
  • VercelCipherfirst sightingNo class

    Vercel: AI SDK harness layer now supports native subscription authentication

    The AI SDK harness layer now supports authenticating harnesses through their native subscriptions, where the underlying harness supports them. The harness layer runs different coding agents through the same HarnessAgent interface, so you can switch agents without changing your application code. No code changes or new settings are required. The direct authentication mode uses explicit provider environment credentials when they are present, and otherwise a native subscription found on the host.

    Read the article →Read the original announcement
  • TurborepoforgeNo class

    Turborepo 2.10

    <!-- Release notes generated using configuration in .github/release.yml at v2.10.13 -->. chore: Release Turborepo 2.10.12 by @github-actions[bot] in. fix: Remove unsupported remote cache environment variable. fix: Copy TUI selections locally over SSH. feat: Use uv workspace metadata. feat: Support Python virtual environments. fix: Scope uv lockfile affectedness. test: Isolate uv prune configuration. fix: Explain disabled uv task caching. fix: Explain uv identity probe failures.

    Read the article →Read the original announcement